Friday, November 13, 2026
Top stories
Jensen Huang resigns; Nvidia names Colette Kress interim CEO
Nvidia confirmed that Jensen Huang resigned as president, CEO, and a board member, effective immediately. CFO Colette Kress will serve as interim CEO while retaining her finance role, and the board is searching for a permanent successor. Nvidia says its customer commitments and product roadmap should continue without interruption.
Why it matters: Leadership continuity at a central AI-chip supplier is now a major operational question for customers and partners—not just investors.
Source: Nvidia’s announcement · November 12
Google launches Gemini 4 Hydrogen for faster agent workflows
Google introduced Gemini 4 Hydrogen for coding and rapid execution, with availability across its developer, enterprise, and consumer products. It also announced an Antigravity orchestrator that switches models within a session. Google reports efficiency gains from this approach, but those benchmark claims should not be treated as guarantees for individual workloads.
Why it matters: Assigning routine execution to a faster model while reserving heavier models for difficult reasoning could make AI automation more economical.
Source: Google’s announcement · November 12
AMD expands its AI accelerator lineup and publishes new inference results
AMD introduced the Instinct MI450X for its Helios platform, offering 288 GB of HBM4 memory per GPU. Separately, AMD reported MI455X throughput of 21,333 tokens per second per GPU on DeepSeek-R1 in an offline inference configuration. These are AMD-reported results, not an independent demonstration of performance across all applications.
Why it matters: More accelerator choices could give infrastructure buyers additional leverage on capacity and cost, provided their software and workloads perform well on the alternatives.
Source: AMD’s announcement · November 12, 8:00 a.m. CST
Advocacy groups ask a court to halt new advanced AI models
Four nonprofit groups sued Anthropic, Google, Meta, OpenAI, and SpaceXSI in San Francisco Superior Court. Their complaint seeks to stop further model development and deployment, alleging unacceptable safety risks under California law. The filing is a request for court intervention—not a ruling that the allegations are proven or that development must stop.
Why it matters: The case opens another route for challenging frontier AI development while broader legislative oversight remains unsettled.
Source: Associated Press reporting · November 12
Australia’s financial regulator prepares AI-risk stress tests
Australia’s prudential regulator plans to begin AI-related stress tests next year, according to ABC’s reporting on its latest risk outlook. APRA warned that banks, insurers, and retirement funds need stronger preparations for disruption and AI-enabled cyberattacks. These are regulatory risk assessments, not a prediction that a particular failure is imminent.
Why it matters: Financial institutions are increasingly being asked to demonstrate recovery and continuity plans, rather than merely describe how they govern AI.
Source: ABC reporting and analysis · November 13 in Australia
From Anthropic
Operate Claude Managed Agents inside your security perimeter
Anthropic introduced a private beta for running its agent infrastructure in a company’s own cloud or on-premises environment. Customers host the agents, credentials, and session history while choosing where requests to Claude are sent.
Source: Anthropic’s post · November 12
J. Crew brings its signature approach to styling to life with Claude
J.Crew launched Style Advisor for U.S. shoppers, recommending complete outfits rather than individual items. Anthropic says the retailer built it with Claude Code and uses different Claude models for different tasks.
Source: Anthropic’s post · November 12
Introducing Claude for Manufacturing
A new plugin provides 17 skills and four connectors for tasks such as drafting quotes, quality reports, and supplier requests. It works with manufacturers’ existing systems and keeps people responsible for reviewing and approving the work.
Source: Anthropic’s post · November 12
Introducing Claude Meetings
Claude Meetings, now in beta, lets Claude join video calls as a visible participant and begin work during the discussion. Teams can ask questions, build shared documents, and use connected tools with permission while Claude records notes and action items.
Source: Anthropic’s post · November 12
Using Claude Opus 5.5 as your daily driver for knowledge work
Anthropic published guidance on using Opus 5.5 for research, analysis, writing, and business documents. The company identifies it as the default model on Pro, Max, Ultra, and Team plans; this is a usage guide, not a new model launch.
Source: Anthropic’s post · November 12
The Claude Tag guide for startups
This guide draws on eight startups’ experiences using Claude Tag in Slack. It explains how teams can bring Claude into channels, connect tools, and assign work together; it is practical guidance rather than a new product announcement.
Source: Anthropic’s post · November 12
Bottom line
Today’s developments point in two directions: vendors are pushing AI deeper into everyday workflows, while leadership uncertainty, litigation, and financial regulators are increasing the pressure for dependable operations. Practical takeaway: test performance claims on your own tasks, limit agent permissions, and keep human approval for consequential actions.
Compiled with AI assistance and reviewed by Dan Kardell.